The pharmaceutical sector faces increasing vulnerability to supply chain disruptions, with significant implications for both healthcare delivery and operational efficiency. Conventional supply chain management frameworks often fall short when confronted with unexpected failures, highlighting the necessity for sophisticated real-time recovery mechanisms. This investigation introduces NEXUS, an artificial intelligence framework that employs Deep Q-Networks (DQNs) to autonomously identify, address and mitigate failures within pharmaceutical supply networks. The framework transforms complex supply chain data into optimized four-dimensional feature vectors that encapsulate critical information regarding feature selection, stability monitoring, constraint identification and anticipatory resilience. By integrating concepts from information theory, statistical divergence assessment and reinforcement learning methodologies, NEXUS facilitates adaptive decision-making that responds to evolving conditions in real-time. Our experimental results demonstrate that NEXUS achieves up to 76% node recovery within 135 steps, with a 93% service level during normal operations and 64% service level maintenance even during black swan events - significantly outperforming traditional approaches. Unlike traditional rule-based approaches, our methodology continuously refines recovery strategies based on historical performance metrics. This paper establishes the mathematical foundations of NEXUS, presents a detailed implementation algorithm and outlines evaluation methodologies across diverse pharmaceutical supply chain scenarios including baseline operations, periodic demand fluctuations, progressive system deterioration, unexpected demand spikes and catastrophic events.